In the news
Nunchi: Ears Help Eyes in Multi-Party Conversations
arXiv cs.AI · Published · 3 min read
In 30 seconds
- What happened
- Nunchi framework improves gaze detection in multi-party conversations by combining visual gaze geometry with speech and transcript cues through vision-language models.
- Why it matters
- Matters for engineers building video understanding systems, meeting analysis tools, or attention-tracking applications that must identify who someone is looking at.
- Watch out
- Requires synchronized speech, transcripts, and speaker identification; gains depend on transcript quality and may not generalize to scenarios lacking reliable audio or speech data.
- speech
- benchmark
The patterns behind this
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.